A human being is affected by "Preconception" by just before memory or impression. There fore, in this paper, a more human consecutive fuzzy reasoning method that considered "Preconception" for the output is proposed. The outputs are given "preconceptions" by this reasoning method. And the next output is adjusted by influence of value of "Preconception". Recognition of a kind of car names in running car images is applied to this reasoning method. By the experimental result, the effectiveness of this reasoning method is examined. As a result of experiments, the consecutive workings of this reasoning method were confirmed.

View full abstract

In this study, a system for supporting human memory is proposed. As a concrete implementation, a person retrieval system accepting fuzzy words is discussed. The proposed system can retrieve a person with fragmentary information the user can remind and with fuzzy words on individual impressions. The individual impressions are automatically extracted from an image using image analysis and are registered as labels about the person. In this paper, the validity of the impression labels extracted by the proposed method is evaluated by subjectivity. The results of the experiments using real data are reported.

View full abstract

This paper describes an image retrieval system aimed at reducing the semantic gap by adopting relevance feedback and learning. To reduce the gap between low-level features and high-level semantics, in this image retrieval system, users'sketches play an important role in relevance feedback. Learning realizes that images are automatically relevance to the sketch images using query-by-example image retrieval. Experimental results show that this method is effective in retrieving some images.

View full abstract

This paper regards on the application of images to information structure education. It is difficult to explain in words only the concept of variables and data structures in computer programs. However, it is possible to do it using words and images. This paper proposes a method for information structure education using images made with spreadsheets.

View full abstract

Although the research accepts the multiple selections to each of the questionnaire,its analysis seems to be almost the same as seen in the one for the usual questionnaire with only one selection. There must be somehow loss of information in the such conventional methods. To this issue do we apply the theory on Indistinguishability Relations in Dempster-Shafer theory of evidence, in which the frequency for all of the subsets will play the main role. In this case the basic probability assignment may be directly determined from the source data. The main concern is in the discussion on the Indistinguishability Relations among the items of questionnaire. The fuzzy relations consist of so called T-indistinguishability operators. Carring out the alpha cuts for them, we obtain a partition concerning the items of questionnaire, which means basically a findings of the essential factors. Some results from the artifitial data and also from the actual ones are presented.

View full abstract

This paper discusses the problems of the similar set retrieval based on averaged error sums with different measures. Since the ranking data of two averaged error sums couldn't be combined for well-informed retrieval, the averaged error grades, which are obtained by normalizing the averaged error sums by the maximum, were employed to get the membership grade of the similar set by the difference to 1. The retrieval system has been designed to find the nearest set based on MIN operation and the ranking of the membership grades.

View full abstract

The music and the picture are thought to be perceived independently, so it is rather difficult to transform the music to the picture and vice versa. Some people insist strongly to imagine picturesque representation from listening the music. It is not impossible to transform the music to picture directly, but the impression of the music can not be represented precisely in the picture by the direct method. We adopt ambiguous transformation and acquire the passable expression.

View full abstract

For the purpose to analytically compare the present meteorological measurement data with the past averaged data, it is proposed to convert the past averaged data to music sounds based on predetermined data-to-sound conversion rules. The resultant music sounds have been recorded on the CD so that the listeners can remind of the past averaged weather of each city. For helping the discrimination of the fuzzy music sounds, the music was clustered into 5 categories based on the relationship between the monthly averaged temperature and the rain fall valume.

View full abstract

This paper introduces the method that analyzes the corporate evaluation in bulletin board sites. The method generates a thread analysis report from the threads included in the sites by using 4 extraction processes, which are article extraction, event extraction, thread extraction, and phenomena extraction. The report is composed of important threads and their relevant information. The analysts can easily check whether the threads are important or not by looking through the report. The paper discusses the functions and the specification in order to construct the practical system incorporating into the method.

View full abstract

This paper reports an application of similar-data retrieval system using case based reasoning (CBR) in petroleum plant. The output of similar-data retrieval system is used in order to identify product properties in crude oil switching. The operational procedure of data-search in crude oil switching is simulated and implemented to the similar-data retrieval. To evaluate, the value of a certain product property by the retrieval system is compared with real operational value. Good results are obtained in the evaluation.

View full abstract

In a petroleum refining plant with a large number of high-pressure facilities, high-pressure gas leaks resulting from equipment failures may engender disasters. To prevent these disasters, technologies for the early detection of leak sound and the appropriate countermeasure are indispensable. We proposed chaos information criteria based on a trajectory parallel measure to analyze the dynamics of acoustic time series data, and reported effectiveness of chaos information criteria on high-pressure gas leak detection, at our previous work. This paper reports that the proof experiment is carried out at Idemitsu Kosan Chiba refinery. Nitrogen gas is artificially leaked at nine different places, we analyze acoustic time series data observed by eight microphones installed in different places. As a result, gas leak detection is possible regardless of the position of a microphone.

View full abstract

We applied fractal-fuzzy control for combustion control in a fluidized bed waste incinerator. Fractal dimensions for the time series monitor image data of incinerator state were sequentially calculated to measure the fluctuation of the data. We applied fuzzy rules using the fractal dimension size to control the amount of secondary air quantity, and the amount of waste feed quantity. Using the proposed fractal-fuzzy control we were able to improve the combustion state.

View full abstract

With the dawn of the fuzzy set theory, many researchers and their fellows made great devotion to the development of applications. For years,they have been highly appraised and implemented inmany fruits-full products. Nowadays, no one will dispute how to use of "fuzzy" as the old problem. However are you clear we can do and can't do with the fuuzy technology?

View full abstract

This paper describes the support system for story creation using some pictures. The input to the system is information on some pictures. The output of the system is incomplete story and descriptions explaining behavior of the object in the picture and connection between pictures for story creation. The user modifies incomplete story to complete the story user feels interesting by using outputted descriptions or adding new description. This paper also shows the experimental results for the confirmation of the usefulness of the present system.

View full abstract

It is a kind of intelligent behavior for human to freely walk and safely drive a car even under complex environments. Our research addresses to the imitation of human action intelligence for the implementation of autonomous mobile robots. Based on the verified effectiveness of a proposed imitation method by means of simulator, we shall further investigate the effectiveness of this imitation method on practical robots. Therefore, we developed a tele-operation system for intelligence imitation robot, including mobile robot and tele-operation interface, then discussed the imitation performance by experiment.

View full abstract

This paper proposes the university guidance system with facial expressions consisting of the emotion setting section and the facial expression section. The former section obtains degrees of 6 kinds of emotions considering user's evaluation of a university and importance grade of attributes for university evaluation. The fuzzy reasoning method and fuzzy measures are applied to this section. The latter part makes facial expressions using degrees of 6 kinds of emotions with the neural network model. The present system gives a user facial expressions reflecting user's evaluation of a university. The usefulness of the present system is confirmed by subjects experiments.

View full abstract

Agents select random action about first element, which come up a view area of agent. The proposed learning system has a multi-layer style rule base. If one rule base has no valid rule, other rule is able to output the incomplete actions by slightly valid elements of the rule base. In any cases, agents should take a behavior by the rule system. The proposed Q-learning system continues learning and adds the new element and changes the next layer when the agents has observed new elements. In dynamic cases, agents should work any cases, but old rules as often as dose not work a current state, because new elements are changed any situations for whole agents, dynamically. In this paper, we discuss the reasonable interrelation level about those rules.

View full abstract

Computer Shogi programs are still in developing to defeat the professional human players. One of the main problems exists in estimating the status of the game phases. It is said that there are three phases in the Shogi game, so called, opening, middle and endgame phase. The appropriate strategy to be selected differs depending on the proceeding game phase. In the present paper, we have carried out the classification of the phases using Self-Organizing Maps and Fuzzy c-Means. The results show the promising feature of the proposed method.

View full abstract

For automatic image annotation, a method based on rough sets with visual keys is proposed. Using rough set theory the method constructs decision rules about each visual key used for image indexing and about keywords from training set of already annotated images. Then target image is annotated according to constructed decision rules about visual keys which the target image is indexed by. The method is evaluated with training sets of 900 images and with test sets of 100 images on 1,000 manually annotated images in COREL database. Experiments show that recall rates tend to rise easily compared with precision rates on image retrieval with query-by-keywords.

View full abstract

A comprehensive comparison of the three variants of region-based image similarity models employing the Earth Mover's Distance (EMD) is performed. An important characteristic of the EMD - that makes it particularly suitable for the region-based image similarity modeling- is that it allows for comparison of images with different number of regions. Despite this fact, very few of the existing region-based models employ EMD. Common for the proposed three variants are the image features and the employment of fuzzy aggregation operators, while the difference is in the way elements of the EMD - flow matrix and feature distributions - are modeled. Empirical evaluation of the retrieval performance of the three variants on four test databases, containing 4444 general-purpose images in 150 semantic categories is done. Comparison of the proposed variants to the representative of the existing region based models is done, too. Results suggest, that proposed models result in different retrieval performance and suggest possible improvements that can be achieved along the same line of research directions.

View full abstract

This paper proposes a method to generate an "intersting" summary from a story and shows a system based on the mothod. The system consists of two sections. The one is the section extracting an important part of a story, i.e., the part of a new turn of a story defined by the appearance frequency of new words in a story. The other is the section generating a summary from the extracted important part. This section chooses important sentences and deletes unnecessary phrases from the extracted part. Finally, this paper confirms the validity of the presented approach by subjects experiments.

View full abstract

This paper proposes the generation system of Japanese riddles considering funniness in wordplays. The proposed system consists of three sections: the generation section, the obtaining section of funniness, and the selection section of riddles. The first section has words databases and generates some riddles candidates. The second section considers some factors and applies fuzzy measures and integrals to the evaluation of generated riddles candidates. This final section chooses riddles according to the their funniness. This paper shows subjects experiments in order to confirm the validity of the proposed system.

View full abstract

In recent years, there are a variety of information technology tools, like personal computers, mobile phones, and car navigations, in the world. These tools have specific interface. It is difficult for us to understand intuitively how to use. So we aim to natural interaction between human and system. Therefore, with joint visual attention, a system focuses attention to the same thing with a human being, and we get joint understanding. For that, we constructed nonverbal interface by using human natural motion. In this study, we focused on using pointing behavior for a natural motion interface. The experiments, the user pointed out a specific parking space. The system recognize that the user want to park a car there. And the system tries to park a car automatically. Furthermore, we intend to make use of this study in RT middleware.

View full abstract

This paper presents controller design for stabilizing lift force of a bird-like flapping robot. In our previous papers, in order to develop a bird-like flapping robot which can autonomously fly, we analyzed aerodynamic force for flapping movements, and developed a mechanism which can change the center of gravity of the robot to achieve bird-like circular motion. In this paper, we construct mathematical model for flapping motion of the bird-like robot based on lift force which is generated from a wing of the robot and control input which we give to a transmitter for controlling the bird-like robot. Based on the mathematical model, we design the controller for stabilizing lift force of the flapping robot.

View full abstract

The robot application system named RAR(Robot Assisted Rehabilitation) is proposed. RAR is a robot system that supports medical treatment. In this paper, we developed the walking rehabilitation support robot.

View full abstract

On up coming aging society, the numbers of elderly people who live alone are increasing rapidly. We need surveillance system for them, by which their family who lives away from them can monitor their health condition. Several monitoring system has been developed using some home application as sensor device for their dairy activities. However, their sensory data are too poor to derive the detailed health condition of the elderly people. Furthermore, the communication media are not carefully examined, which results that these systems are not used well. This paper presents the surveillance system that monitors the everyday health condition of elderly people in detail. The system is focused on detecting unusual activity. As the first step, we are examining to classify the variety in walking, e.g., limp, walking with head bent forward, etc.

View full abstract

This paper describes a study on electroencephalogram (EEG)-based control. Discrimination methods for left/right direction thinking using brainwaves have been proposed. The main target of these studies were to develop control systems using the recognition result. Effects of measurement locations on human head on the recognition performance were not investigated. This paper employs Linear Discriminant Analysis (LDA) for this discrimination of human thinking. LDA makes us easy to investigate the relationships between the measurement positions and the recognition rates. The authors focus on alpha and beta waves which are called specific frequency range included in brainwaves, and we try to discriminate direction thinkings using the alpha and beta waves. Experimental results with four subjects show that average discrimination rate is approximately 70%. This paper also shows a preliminary investigation of eective measurement positions for the thinking detection.

View full abstract

Recently, researches on brain wave based controls of robots have been reported. They are achieved by thought recognitions using brain waves. The authors have proposed a method that transforms brain waves to frequency series and apply a discriminant analysis to recognize one's thought. The conventional method uses a variable selection by a cross-validation to prevent an over-training. The number of candidates of variables is, however, a product of that of measuring locations times that of frequency series. As a result, it is impossible to calculate all the combinations of them within a real-time. This paper proposes a new variable selection method based on p-values in the t-test. It does not take much time to calculate p-values. In addition, it is expected that the proposed method will improve the recognition accuracy because it can select variables from the candidates that are lost in the conventional method. It is also expected that it makes possible a real-time training and therefore it enables us to feed back information such as fatigue and accuracy to subjects during measurements.

View full abstract

Japanese texts with interpersonal meanings focused on honorific expressions and ideational meanings were presented to four subjects. In the meantime electroencephalograms (EEGs) were recorded, and equivalent current dipoles (ECDs) were localized to estimate the activated neuronal sources of EEGs. The experimental results showed two different dynamical routes in the brain common to all the subjects in understanding honorific expressions in contrast to texts with ideational meanings.

View full abstract

In the human brain, the logical process is performed at the left hemisphere and the intuitional process is performed at the right hemisphere. We recorded event related potentials (ERPs) by the electroencephalograms (EEGs) during subjects were performing selections of an answer. The subjects were 7 normal university students (4 males and 3 females; aged 21-22 years; 6 right-handed and 1 left-handed). They were asked to add double figures and select an answer from the two choices. Choices were two types; Type1 was correct choice that the correct answer was included, and Type2 was approximation choice that the correct answer was not included. Moreover, we classified into following cases according to approximation choice. The case was Type2A: the answer was lower than choices. Type2B: the answer was between choices. Type2C: the answer was higher than choices. By using the equivalent current dipole source localization (ECDL) we compared brain activities in the meantime among above types.

View full abstract

The modified fuzzy c-means (FCM) clustering is used for exploratory analysis of fMRI signals in which the cluster centroids represent typical hemodynamics responses of relatively wide brain regions. This paper shows how blood oxygen level-dependent (BOLD) signals can be approximated by a small number of cluster centroids and presents a framework that allows FCM clustering technique to be implemented in conjunction with the general linear model (GLM). The approach is illustrated with fMRI experiments of recalling sentences. Because the recalling by way of mnemonics is a heavy task for the brain, the wide brain regions activate similarly. This makes it difficult to distinguish the role of brain regions. The exploratory approach clarifies the slightly different responses of different brain regions to the tasks of recalling sentences. The visual areas activate more than the temporal and frontal areas. The visual areas seem to play a significant role in recalling sentences.

View full abstract

Aperture problem is an experiment for anlying the early motion processing in the visual pathway. Nishina argues the perception depends on the display time and discuss the processing of visual computation tasks. In this paper, we measure various perceptual rates in the same experiments, and discuss the dependency between the perception and the moving speed of stimulus lines. We also simultaneously record Electroencephalograms(EEG) of subjects who are recognizing the perception, and then estimate the source and the latencies in the brain.

View full abstract

Online learning machines are essential tool for real-time adaptive systems such as autonomous robots. However, if the input data include a large number of unrelated variables, the learning machine wastes a high computational power to complete the learning. This is because, the learning machine needs a huge number of samples to detect which variables are related to the desired outputs precisely. To overcome this problem, we propose a quick online learning methods using speculative filter and wrapper methods. In this method, the system achieves quick online dimension selection and learning even if the input samples are independent distributed samples.

View full abstract

TAM network is a biologically-motivated neural network with receptive fields of Gabor function. However, it's not good performance for rotating visual images since the receptive field's layer is a mono-type. In this paper, we formulate four types of multilayer structures with extensive receptive fields, and discuss the performance using some examples of character recognition.

View full abstract

In any tele-operation environment, the behavioral system of the "worker" is imposed to physically separate into two parts: the cognitive and the performatory part of the worker. The former is mainly taken on by a human operator while the latter is done by his surrogate performer, i.e., a remote controlled robot. Such a divorce of the behavioral system may disrupt the perceptual-motor couplings and then the coordinations among different subordinate behaviors. This work focused on those cooperation structures between perceptual and motor systems in ambulation, and compared some perceptual evidences of a mobile robot tele-operation with those of naturalistic human ambulatory movements (walking and bike-riding) from a perspective of ecological phsychology. Effects of a purposefully embedded connection between robotic behaviors for viewing and ambulation were examined in terms of the system behaviors and usability.

View full abstract

Recently, social interaction is required for human-friendly robots to communicate with humans.A partner robot requires various capabilities for the social interaction based on verbal communication and non-verbal communication, but the perceptual capabilities is much lower than humans. In this paper, we focus on the communication with robots for the utterance learning where envirtonmental information is avaliable from an environmental system. Finally, we show several preliminary experimental results of the proposed method.

View full abstract

The patterns of spontaneous action potentials are analyzed using the multi-site recording system for the living neuronal networks cultured on a 2-dimensional electrode arrays. In this paper, we constructed ``biomodeling system'' for Khepera II robot and living neuronal network through fuzzy logic. Our goal is reconstruction of the neuronal network, which can process ``thinking'' in the dissociated culture system.

View full abstract

In binocular vision, observers can perceive transparent surfaces by fusing the stereograms generated by overlapping two dot patterns in different depths. In the case that overlapping patterns are identical, the stereogram has the matching candidates that can generate both transparent and non-transparent (or unitary) surfaces. Here we investigate the transparency perception in this ambiguous stereogram with additional segragation cues for overlapping surfaces. We also show the population coding model with the hybrid disparity energy model can explain the psychophysical result.

View full abstract

It has been reported in many papers that the number of eye movements increases and the mean fixation durations on individual objects decrease when the road situation becomes more complex. In this paper eye-movement analysis has been used for understanding mechanism of visual search of drivers. For the aim, an experiment in which the subjects drove the driving simulators have been conducted. We discuss the influences of the peripheral vision for the eye movements characterized by averaged frequency of saccadic eye movements and fixation duration.

View full abstract

Recently, the concern is paid to attention by the entertainment computing that is the amusement with a computer. The availability of the Internet increases with the spread of a rapid personal computer from the latter half of the 1990's, and the convenience of life has gone up greatly. In this paper, we propose a new method of using IC tag about an automatic composition of music by the computer. The entertainment concerning the automatic composition that we think about is a field where man's inside or piece can be improved more. We propose the music composition support system that makes the tune by "Dynamic" method different from a past "Static" method.

View full abstract

The interactive sound is one of the interactive art researches in the artificial life field. The research purpose of interactive sound is to achieve the sound with variety and complexity, which exceeds the forecast of human, based on the interaction between human and a system. By using the network of chaotic elements controlled both chaotic behavior and whole synchronism, sounds with various aspects are generated. In this research, some parts of musical element are also included, and the incorporation of amusement effect used the image information by which human can interactively take part is attempted. We developed ICAS(Interactive Chaotic Amusement System) and performed Kansei evaluation by the questionnaire for several sounds generated by this system to confirm the amusement effect.

View full abstract

The communication tools increase in the web recently, it becomes more and more important. For example, communications has not only content of word but a lot of information of the expression, the voice and the body language. It can be more smooth communications by telling nuances and vagueness. But how is in the web society? It is chiefly communications only of the type data on the web. An inefficiency element is deleted from our communications, and information is abstracted only to the character. There is danger to which a little word gives birth to an unexpected friction though it is efficient. It is a purpose of this research to facilitate the feelings transmission by using Bayesian Filtering and Ekman's thought.

View full abstract

Various designs exist in the world now. Especially, in the net society, we have an interest some designs such as character design, web page design, and logo design, etc. In these designs, the requested element is different respectively. The T-shirt design has various elements though it is simple. The T-shirt design is one of the means to insist on individuality. It tends for the one unique as for the dress to be requested in recent years. In this paper, we propose a method of the T-shirt design that is taken up in the evaluation object, which is evaluated by the heavy regression analysis.

View full abstract

This paper describes the construction of a system which transforms a theme music fitting to story scenes represented by texts and/or pictures, and generates variations on the theme music. Inputs to the proposal system are an original theme music and numerical information on given story scenes. The present system varies (1) melodies, (2) tempos, (3) tonalities, and (4)accompaniments of given theme music based on impressions of story scenes. Neural network models are applied to the music generation in order to reflect user's sensitivity on music and stories. This paper also describes the evaluation experiments to confirm whether the generated variations on theme music reflect impressions of generated story scenes appropriately or not.

View full abstract

Emotion expression through behavior is meaningful for communication processes between robots and humans. Automatic emotion expression generation is needed for those robots in order to communicate with humans smoothly. Proposed technique can generate emotional motions by modifying base motion patterns applying adjectival modification rules. Adjectival modification rules are consisted of the relationships between emotions and adjectival expressions. In this paper, we perfirmed organoleptic test in order to confirm that proposed technique is applied to vaious base motions. And this paper shows a major difficulty of applying adjectival modification rules to base motions with high degree of freedom based on the result of the organoleptic test.

View full abstract

In this paper, we develop FCM relation rules using a "concept template" instead of a concrete concept description. The concept template consists of a predicate and the case elements of the predicate, and they are used for both input concept and output concept in relation rules. The input concept templates accept the variation of similar input concepts and the output concept templates are used to generate the content of the output concept based on the content of input concepts. The inference process will be repeated until the concept state reaches convergence. We created 207 concept templates and a weighted matrix to show the relationship between concepts. When 23 concepts were inputted to our system, 71 concepts were totally activated and the content of all output concepts were generated correctly.

View full abstract

In this research, we try to develop a Kansei robot which can recognize color combination image that means "feelings about colors". In that robot, the color image evaluation system, which has developed in our previous research, is installed. By using the system the robot can evaluate feelings about some color combinations and express these impressions in various kinds of key words, such as pretty, casual and chic. This paper describes to develop a Kansei robot recognizing feelings about colors of clothing which a person puts on and expressing these feelings in some key words.

View full abstract

In this paper, we propose a target selection-type Q-Learning method with plural Q-values concerning the maximization and minimization of rewards and punishments. We aim at the realization of a system to obtain complicated emotion behaviors selected based on the positive and negative evaluation according to the situation. Furthermore, we also report the results of an experiment by computer simulation and Kansei evaluation to confirm the efficiency of proposed method. A new simulation to show the effect of an internal state is also introduced at the end of this paper.

View full abstract

This paper describes new Laban's Feature Values for dog robot. A classical approach to feeling expression using Laban's Feature Values was just for human type robot. We applied the focus to the movement of the tail of dogs when they express pleasure. Then, new feature "Rolling Effort" is proposed. In experimental results using AIBO, the effectiveness of new Laban's Feature Values include "Rolling Effort" are shown.

View full abstract

This paper proposes the evaluation system that considers the influence of information on the poster design gives us impression. First we carried out a questionnaire about impressions of posters. Second we measured the elements included in posters. Here, a multiple regression analysis was done and the forecast value and the coefficients of determination value were obtained. We analyzed it for the subject's classification dividing into a man and the woman. As a result, the items that have differences in man or woman became clear, and the necessity of the classification of the subject was recognized clearly.

View full abstract